Scribd
Upload
14 views

Objection detection

The document outlines a project aimed at developing a real-time object detection system using audio assistance for visually impaired individuals. It discusses the challenges faced by existing systems, the proposed solution utilizing machine learning and computer vision techniques, and the expected outcomes of improved object recognition and navigation. The project includes various components such as system architecture, experimental results, and individual contributions from team members.

Uploaded by

dhinahari.raja
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
14 views

Objection detection

The document outlines a project aimed at developing a real-time object detection system using audio assistance for visually impaired individuals. It discusses the challenges faced by existing systems, the proposed solution utilizing machine learning and computer vision techniques, and the expected outcomes of improved object recognition and navigation. The project includes various components such as system architecture, experimental results, and individual contributions from team members.

Uploaded by

dhinahari.raja
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 25

UNIVERSITY COLLEGE OF ENGINEERING

BIT CAMPUS
ANNA UNIVERSITY, TIRUCHIRAPALLI-620 024
CS8811 - Final Project

ENABLING OBJECT DETECTION THROUGH


SPEECH FOR VISUALLY IMPAIRED
Under the Guidance of Project by
N.Shanmugapriya,ME Dhinahari R-810020104023
Teaching fellow Dhivya A S-810020104024
UCE-BIT CAMPUS BATCH NO-26
Trichy-24
AGENDA
Introduction Algorithm and Technique
Objective System Requirements
Literature Survey DFD Diagram
Existing System Experimental Results
Proposed System Gantt Chart
System Architecture Contribution
Graphical Abstract Conclusion
Module Description Reference
INTRODUCTION
Object detection for visually impaired individuals using audio assistance
represents a transformative advancement in recent technologies.
Visually impaired individuals face challenges in understanding the
environment and may experience difficulties with navigation and social
interaction.
Limited applications have been come in market to aid the visually impaired
people in understanding the environment.
Realtime object detection increases the freedom and safety of visually
impaired people by allowing them to explore and interact with their
surroundings more confidently with the help of voice assistance.
OBJECTIVE
To develop a Real-time object detection system using audio assistance for
visually impaired to recognize objects independently.
To utilize computer vision techniques and machine learning algorithms for
detecting objects and persons.
To convert text into voice response after identifying the object.
LITERATURE SURVEY

SI.N
O
TITLE YEAR AUTHOR METHOD LIMITATIONS

An Evaluation of RetinaNet on
Challenging to capture
Indoor Object Detection for Blind
1 2020 Mouna and Riadh CNN and identify the objects
and Visually Impaired Persons
with high accuracy
Assistance Navigation

Due to complexity, multi


HCP: A Flexible Convolutional
labelling is an open
2 Neural Network (CNN) Framework 2015 Wei and Xia CNN
challenge for training
for Multi-label Image Classification
image layouts

Indoor Object Recognition in a


combination of a RGB image and
Takes more time for
3 its corresponding depth image 2018 Rim and Issam CVNN
detection
(RGBD) Images for Visually-
Impaired People
EXISTING SYSTEM
Limited Object Recognition Accuracy: Many existing systems struggle with
accurately identifying objects, particularly in complex environments.
Latency Issues: Some systems may experience latency or delays in object
detection and recognition, especially in real-time applications.
Dependency on Environmental Factors: Certain systems are highly dependent
on environmental factors such as lighting conditions, object placement.
Restricted Vocabulary: Some systems may have a predefined vocabulary or
limited database of recognized objects, which can restrict the system's ability to
identify a wide range of objects encountered in daily life.
PROPOSED SYSTEM
The proposed system integrates machine learning techniques to develop a
comprehensive real-time object detection system.
Preprocessing the image using different methods like cropping, resizing,
filtering and enhancing the captured images and clear the background
without any obstacles.
Using machine learning algorithms to map a object viewed with objects which
are in dataset.
After matching is done, text-to-speech conversion(TTS) to create speech
synthesized from the translated text.
ARCHITECTURE DIAGRAM
Testing phase
Training phase
Train models KNN Algorithn

Evaluate results

Sample
Precision
images
Feature Extraction Recall
mAP

Classification

Live phase

Webcam input

Audio output
GRAPHICAL ABSTRACT

API
Get camera Object convert object
feed detection name into speech

Input

Object detection
algorithm
Dataset
Audio output
MODULE DESCRIPTION
Data Collection: Continuously collect image samples in real-time through the
webcam. Label the object in these data samples.
Incremental Training: Incorporate the newly collected data into the model and
update its parameters periodically.
Real-time Detection: Use the updated model to detect objects in real-time video
streams on webcam. Apply the detection algorithm to each frame or image as it
becomes available, and display the results in real-time.
Adaptation: Monitor the performance of the model over time and adjust its
training strategy or architecture as needed.
ALGORITHM
KNN ALGORITHM
KNN Algorithm classifies unknown data points by compairing unknown data
point to each data point in training set.
It is machine learning algorithm used for classification and regression.
It involves data preparation, feature extraction, training, detecion, classification,
post-processing.

Find the k nearest


Read value of k, type of Set maximum label
START neighbours D to the END
distance (D) & test data class of K to test data
test data
SYSTEM REQUIREMENTS
Hardware Specifications
Processor: Dual-core processor or higher
RAM: 4GB or more
Storage: Minimum 20GB of free disk space
Internet Connectivity: Required for accessing online features and updates

Software Specifications
Operating System: Windows 7/8/10/11
Programming language: html, css
Scripting language: JavaScript
Web browser: Google Chrome, Opera, Firefox
DATAFLOW DIAGRAM

LEVEL 0

VIDEO INPUT MATCHED OBJECT


WEBCAM OBJECT DETECTION AUDIO OUTPUT
LEVEL 1

OBJECT DETECTED
WEBCAM
DETECTION OBJECT DATA

OUTPUT TRACKED OBJECT


OBJECT DATA TRACKING
LEVEL 2
DEEP
TRAINING EXPORTED
PREPROCESSING LEARNING
DATASET MODEL
MODEL(KNN)

DETECTED
WEBCAM OBJECT DETECTION
OBJECT DATA

MATCHED FINDING
OUTPUT
OBJECT MATCH
EXPERIMENTAL RESULTS
EXPERIMENTAL RESULTS
EXPERIMENTAL RESULTS
EXPERIMENTAL RESULTS
EXPERIMENTAL RESULTS
GANTT CHART
INDIVIDUAL CONTRIBUTION

DHINAHARI R Data Collection


Preprocessing
Model selection and development
Training and evaluation

DHIVYA A S Deployment
Testing and validation
Documentation
Maintanance
CONCLUSION
In this project, we developed a real time object detection aimed at assisting
visually impaired individuals through auditory feedback.
We achieved promising results in detecting objects in the surrounding
environment and translating them into meaningful audio output.
Through extensive experimentation and evaluation we observed that the
KNN Algorithm provided satisfactory accuracy and performance.
The integration of audio feedback proved to be an effective approach for
conveying object information to visually impaired users in realtime.
REFERENCES
[1] Xing Yang, Yunsong Feng, Yuanpu Zhou (The ObjectDetection Based on Deep
Learning,doi.org/10.1109/ICISCE.2017.156)
[2] Md. Moshiur Rahman (Real-Time Object Detection using Machine Learning,
doi.org/10.1109/ICCCNT51525.2021.9580170)
[3] Argel A. Bandala, Reagan L. Galvez (Object DetectionUsing Convolutional Neural Networks,
doi.org/10.1109/TENCON.2018.8650517)
[4] Aishwarya Sarkale, Kaiwant Shah, AnandhjiChaudary,Tatwadarshi Nagarhalli(An Innovative
Machine Learning Approach for Object Detection and Recognition,
doi.org/10.1109/ICICCT.2018.8473221)
[5] Ayoub Benali Amjoud, Mustafa Amruch(ObjectDetection Using Deep Learning, CNNs and Vision
Transformers: A Review, doi.org/10.1109/ACCESS.2023.3266093

You might also like